Research map / Adversarial robustness and security / Attacks on text and graphs
MCP Security Bench (MSB): Benchmarking Attacks Against Model Context Protocol in LLM Agents
Dongsen Zhang, Zekun Li, Xu Luo, Xuannan Liu, Pei Pei Li, Wenjun Xu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| combine PI and UI attacks into a mixed attack | LLM agents (avg of 10) · MSB | ASR: PI 20.21, UI 45.69 -> PI-UI 56.07 | Improved |
| use user-impersonating tool response (UI) instead of false error (FE) | LLM agents (avg of 10) · MSB | ASR: 39.21 -> 45.69 | Improved |
| add MCIP detection-based defense to MCP agents | LLM agents (avg of 10) · MSB | ASR: 40.35 -> 28.69 | Improved |
| add MCIP defense to MCP agents | LLM agents (avg of 10) · MSB | PUA: 56.50 -> 48.91 | Hurt |
| add MCIP defense to MCP agents | LLM agents (avg of 10) · MSB | NRP: 33.70 -> 34.88 | No clear effect |
| add MCIP defense to GPT-5 agent | GPT-5 · MSB | NRP: 52.99 -> 46.73 | Hurt |
| use more capable LLM backbone (higher PUA) | LLM agents · MSB | ASR: DeepSeek-V3.1 ASR 60.94 vs Llama3.1 8B 19.74 | Hurt |
Closest papers by meaning
- MCPTox: A Benchmark for Tool Poisoning on Real-World MCP Servers (AAAI 2026 · similarity 0.92)
- Agent Security Bench (ASB): Formalizing and Benchmarking Attacks and Defenses in LLM-based Agents (ICLR 2025 · similarity 0.91)
- Attractive Metadata Attack: Inducing LLM Agents to Invoke Malicious Tools (NeurIPS 2025 · similarity 0.91)
- SEC-bench: Automated Benchmarking of LLM Agents on Real-World Software Security Tasks (NeurIPS 2025 · similarity 0.90)
- MCP-Atlas: A Large-Scale Benchmark for Tool-Use Competency with Real MCP Servers (NeurIPS 2026 · similarity 0.90)
- Breaking Agent Backbones: Evaluating the Security of Backbone LLMs in AI Agents (ICLR 2026 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
